\begin{table}[h!]
    \centering
    \caption{Bootstrap Estimates of Citywide Supply Response and Marginal Fiscal Cost}
    \label{tab:bootstrap_ests}
    \begin{tabular*}{\textwidth}{@{\extracolsep{\fill}}lcc@{}}
\toprule  \addlinespace
                     & Exogenous Construction                & Endogenous  Construction               \\    \addlinespace \cmidrule{2-3} 
                     & (1)                        & (2)                         \\   \midrule \addlinespace
Supply Response    & $$elasticity_estimate$$***                        & $$elast_gmm_estimate$$***                           \\
                     & ($$elasticity_stderr$$)                      & ($$elast_gmm_stderr$$)                           \\  \addlinespace
Marginal Fiscal Cost & \$$$mcost_estimate$$***                        & \$$$mcost_gmm_estimate$$***                              \\
                     & ($$mcost_stderr$$)                      & ($$mcost_gmm_stderr$$)                      \\  \addlinespace  
Average Fiscal Cost & \$$$acost_estimate$$***                        & \$$$acost_gmm_estimate$$***                              \\
                     & ($$acost_stderr$$)                      & ($$acost_gmm_stderr$$)                      \\  \addlinespace \bottomrule
\end{tabular*}
	\begin{spacing}{1}
\begin{tablenotes}
      \item \footnotesize \textit{Notes:} This table presents citywide supply responses and citywide average and marginal fiscal costs. In the first row, I report estimates of the citywide supply response of onsite inclusionary units to changes in buildings' 421-a incentive. In the second and third rows, I report marginal and average fiscal costs per inclusionary unit per year. In the left column, my estimates are from the baseline approach, which assumes building features are exogenously determined. In the right column, my estimates are from the simulated-instruments GMM approach. The original specification includes fixed effects for borough and year as well as lot and block controls. Standard errors are computed by a cluster-bootstrap at the level of Neighborhood Tabulation Areas. $\sym{*} = p < 0.10$, $\sym{**} = p < 0.05$, $\sym{***} = p < 0.01$.
      \end{tablenotes}
      \end{spacing}
\end{table}